In order to preliminarily gauge the comprehensiveness of our data, we compared the records contained in AidData's Chinese Official Finance to Africa Dataset, Version 1.0 with four existing data sources of Chinese official finance. First, to determine the extent to which our data match the (admittedly limited) data on Chinese aid from official sources, we cross-checked our project records with the project records reported in China's MOFCOM Yearbooks from 2000-2005 (with the exception of 2002 when no data were reported).50 Matching our data to MOFCOM Yearbooks proved difficult, as the Yearbooks report project completion years while our database records project commitment years and then follows up on whether projects have been implemented and/or completed. As such this is was a highly imperfect matching exercise. That said, the results from the matching exercise suggest that our database contains more projects listed in MOFCOM Yearbooks for more recent years. This makes sense because commitment years for earlier projects have a higher probability of occurring before 2000—our data collection cut-off date. We matched 6% of MOFCOM projects completed in 2000, 27% in 2001, 50% in 2003, 62% in 2004, and 50% in 2005. This excludes cases in which not enough information was available to discern whether a match existed.
Second, we cross-checked our database with humanitarian aid data recorded in the Financial Tracking Service (FTS). Managed by the UN Office for Coordination of Humanitarian Affairs (OCHA), FTS data are provided by donors and/or recipient organizations.51 It appears that our database contains substantially more reported Chinese humanitarian assistance activities in Africa than FTS for the period 2000-2011. FTS contains 26 humanitarian assistance project records that would plausibly meet our database inclusion criteria. These are cases of Chinese assistance to Africa that fall within the 2000-2011 time range. Of these 26 records, there are 7 for which the available information is insufficient to determine whether or not a match exists between our dataset and the data contained in FTS.
Of the remaining 19 FTS records, 13 (68%) can be matched to a specific project in our dataset. While our data do not match up perfectly to FTS, the evidence suggests that we are collecting more comprehensive and detailed Chinese humanitarian assistance data than FTS.
Our dataset contains 86 official finance projects coded as “Developmental Food Aid/Food Security Assistance” and “Emergency response.”
Third, we have compared AidData's Chinese Official Finance to Africa Dataset, Version 1.0 with the Food Aid Information System (FAIS), an online database provided by the UN World Food Programme (WFP) that tracks international food aid flows.52 Results were mixed. On one hand, we found that FAIS reported over 40 recipient-year pairings with food aid from
50 Data are available at http://www.aiddata.org/content/China-foreign-aid.
51 Data are available at the OCHA website. See http://fts.unocha.org/.
52 Data are available at http://www.wfp.org/fais/.
China that did not exist in our database. But we also found 10 pairings in our dataset that were not in the FAIS database. There were over 10 pairings that showed up in both
databases. However, there are two important disclaimers to be made about this comparison.
First, similar to FTS, FAIS tracks completed projects, in the form of food aid deliveries. Our dataset starts with project announcement dates. Thus, while food aid projects are more likely to be completed in the same year they are announced, we are, in a sense, making apples-to-oranges comparisons.53 Second, FAIS does not provide data for 2010 and also only reports Chinese food aid to 30 African states, excluding a substantial number of recipients for which AidData has food aid records. The AidData-FAIS matching results suggest that our
methodology may not be as effective for collecting food aid data as it is for tracking Chinese foreign aid in other sectors. But FAIS also seems to suffer from substantial data gaps in reporting Chinese food aid to African countries since 2000. Taken together, these
comparisons with MOFCOM Yearbooks, FTS and FAIS suggest that media-based data are no substitute for official data but a viable second-best solution, particularly when official data are largely incomplete.
Fourth, we cross-checked a database of incoming aid flows managed by Malawi's Ministry of Finance. Malawi’s Aid Management Platform contains data from 30 donor agencies and US
$5.3 billion in commitments (current USD), representing approximately 80% of all external funding reported to the Ministry of Finance since 2000. Out of 2584 projects in the AMP Malawi database, only two records (2008 and 2009 project) list the People's Republic of China as the donor entity, totaling $163 million (current USD). Both of these projects are included in AidData's Chinese Official Finance to Africa Dataset, Version 1.0.54 However, our dataset includes 14 additional Chinese official finance projects in Malawi, totaling US$ 164.8 million in commitments. Collectively, these projects double the amount of recorded
commitments of Chinese official finance in Malawi. This cross-checking exercise not only calls attention to the incomplete nature of the data in Malawi’s Aid Management Platform, but also to the fact that donors that do not publish project-level data, such as China, are likely responsible for a substantial proportion of unreported external funds flowing into Malawi. This comparison illustrates the added value of using MBDC as another method to track aid flows in the absence of official project records.
In addition to comparisons with these four official databases, we compare the annual amount of total Chinese aid to Africa, as represented by AidData’s media-based data and estimates from previous studies (see again Table 1). AidData's Chinese Official Finance to Africa Dataset, Version 1.0 contains 937 "ODA-like" project IDs with an aggregate value of US$
53 In our dataset, 52% of official finance projects in sectors “Developmental Food Aid,” “Emergency Response,” and“Agriculture, Forestry and Fishing,” “ have a reported status of "completed," while only 43% of active projects in the entire database have a reported status of "completed."
54 The financial value of one of these two projects, the construction of a hotel and business center, differs between records in MBDC China and AMP Malawi. The former reports a value of $92.3 million, while the latter reports a commitment worth $63 million (and a cumulative disbursement of $80.16 million; all values in constant 2009 US$).
13.0 billion (in constant 2009 US$). The 937 figure includes projects identified as being in the “Commitment,” “Implementation,” or “Completion” stages, and excludes projects with a status of “Pledge.” This is an average of less than US$ 1.1 billion of Chinese ODA to Africa per annum during the twelve year study range. This is roughly comparable to previous studies such as Bräutigam (2011b), Wang (2007) and The Economist (2004) that estimated Chinese ODA to Africa to be somewhere between US$ 1 and US$ 2 billion for a particular year in our study's time range. Additionally, since this number does not include those projects for which we did not find information that they have reached the commitment stage, it is possible that we are underestimating the actual amount of 21st-century Chinese ODA to Africa since some of these projects may have actually been carried out. More broadly, our database contains 1,422 projects that have been classified as “Chinese Official Finance,” which includes projects labeled as "ODA-like," "OOF-like" and "Vague Official Finance," for a total of US$ 75.4 billion between 2000-2011, or US$ 6.3 billion per year. This estimate falls in between previous wide-ranging estimates such as the CRS/NYU Wagner School study that placed 2007 Chinese "aid and related activities" at US$ 18.0 billion (Lum et al. 2009), and Christensen (2010), who estimated 2009 Chinese "aid" to Africa at US$ 2.1 billion.
AidData’s aggregate estimates must be considered in light of two important caveats. First, our estimates not only include data for completed Chinese aid projects, but also for projects in the “Commitment” stage that have been announced or remain in the preparation/design phase but have not necessarily broken ground, as well as for projects for which
implementation is underway but that have not been reported as completed. The total values for Chinese official finance are considerably smaller when we exclude projects that lack information that they have been finalized (US$ 19.4 billion over the 2000-2011 period) or have at least been started (US$ 48.6 billion). AidData’s online data platform at
china.aiddata.org allows users to filter projects and generate aggregate statistics based on the status of a project. Second, 38% of the official finance records in our database lack financial values. It therefore stands to reason that we may have under-estimated Chinese official development flows to Africa in this paper as a result. We hope to fill in as many of these missing financial values as possible in future updates to the dataset.55 To obtain more accurate estimations of the total monetary value of China’s development finance, future research should elaborate ways to impute missing monetary values of individual projects based on their observed characteristics.
55 The previously described web-based platform that allows feedback on projects from recipient governments, journalists, scholars, and other stakeholders is one potential source of information on this and other fields in the database.
44 8. Conclusions and Next Steps
There is a growing disconnect between the suppliers of global development finance and the international regime put in place by sovereign governments to track development finance activities. While the member states of the OECD-DAC by and large comply with a basic set of data disclosure norms, important non-DAC donors have effectively opted out of the global aid reporting regime. Left unattended, this gap will continue to grow. As some Western governments scale back their development finance commitments, non-Western donors are rapidly expanding their overseas aid activities. The most important provider of official finance to Africa among these DAC donor countries is China. Yet many non-DAC donors, including China, lack either the capacity or the political will to provide detailed information about their aid activities. The global aid reporting regime faces a crisis of relevance and legitimacy, and these cracks in the foundation of a voluntary disclosure system developed more than 50 years ago pose a major challenge to scholars and policymakers who seek to understand the distribution and impact of development finance. This paper is the first in a series of efforts to track non-DAC development finance through the application of AidData’s MBDC methodology. We have created a public good that we hope will be used—
and improved—by researchers, policymakers, and other interested stakeholders to better understand the rapidly expanding field of non-DAC development finance.
Apart from contributing to the literature on Chinese aid, we pursued this project as a proof of concept exercise to test the viability of a media-based data collection approach. We regard this pilot project as a success. The methodology has shortcomings and will no doubt be improved, but its application has uncovered more than US$ 75 billion in commitments of official Chinese financing flows to Africa that were previously unrecorded—in a single location and with a single, consistent methodology—at the project level. We hope that this database will be used by scholars, policy analysts, journalists, and others to address important policy questions about the distribution and impact of Chinese aid to Africa. However, we also hope that we have demonstrated media-based methods can substantially increase the transparency of aid flows from Iran, Saudi Arabia, Venezuela, Cuba, and many other donors that are not part of the OECD-DAC reporting regime.
Based on insights from previous initiatives tracking Chinese aid and investment flows, we have taken steps to avoid pitfalls of relying on public media reports by crafting a systematic, transparent and replicable methodology and database. All projects in the database are tracked closely over time, and we have taken extra caution to avoid double-counting of projects. We have also attempted to categorize and present our data in a way that enables analysts to include and exclude certain financing flow types, depending on the nature of their inquiry.
We harbor no illusions that we will definitively resolve the debate about how to categorize different forms of Chinese (development) finance, but we hope that by disclosing our data and methods we will facilitate productive discussion and perhaps make a modest
contribution to the advancement of social science and evidence-based policymaking. Our data collection methodology is publicly available at
http://china.aiddata.org/MBDC_codebook. We also encourage users of our database to
scrutinize the data, and provide feedback and alternative sources of information. On the china.aiddata.org platform, users can access a live, interactive version of the database and suggest new sources of information for—or specific changes to—any project record. Users can also add records if they have knowledge, and the corresponding sources, to verify that a Chinese project has been pledged, committed, implemented, and/or completed that is not contained in AidData’s MBDC China database.
Going forward, we intend to continuously update project records in our database based on user feedback, and update our China database to account for development finance flows beyond 2011. Additionally, we plan to improve and expand upon our MBDC data collection efforts, including the China dataset, in the following ways:
1. Vet and refine project records through correspondence with knowledgeable local stakeholders in Africa.
Media-based data are no substitute for official data. But official data also suffer from a set of known shortcomings—e.g., project-level disbursement and implementation information is often missing or inconsistently reported (Strange et al. 2013). To this end, AidData is exploring a range of options for collecting data from policy-makers, development
practitioners, journalists, and other local stakeholders in Africa who can vet and enhance the Chinese development finance data with insights from the field. This is in addition to our crowdsourcing platform described above. Our first attempt to crowd-source Chinese development finance data is a dynamic data platform (china.aiddata.org) which allows users to investigate and suggest revisions to individual projects. By searching and filtering through the online data or inputting the unique project ID number, users can access the project page, which includes links to source documents as well as a list of contacts who have some knowledge about the project. The project pages also provide a comment function, where users may offer additional project information, link and upload new project documents, or report potential errors. AidData staff will track and moderate these comments, addressing data quality issues as they arise and integrating verified content into the database. Greater participation by local stakeholders will add tremendous value to this process.
2. Expand Stage One and Stage Two to include more searches in additional languages.
Due to resource constraints, Stage One of this pilot project was carried out entirely in English. Factiva also includes rich media databases in many other languages including Mandarin, French and Portuguese that may potentially yield additional projects and/or richer details for existing projects. During Stage Two a team of three Chinese language experts located Chinese sources for aid projects that were initially identified from non-donor and non-recipient news agencies. However, because of resource limitations it did not utilize language searches in languages other than English and Mandarin. In future iterations of our
data, AidData’s MBDC methodology will include more diverse language searching throughout Stage One and Stage Two.56
3. Geocode the precise latitude and longitude coordinates of all projects and analyze the spatial distribution of Chinese development finance.
Later this year AidData will release an updated Chinese development finance database with subnational geocodes. These data will help address a range of questions, such as the degree to which Chinese aid effectively targets areas of need or opportunity and whether Chinese aid is used to curry favor with African political leaders. Among many other applications, researchers can pair geocoded Chinese aid information with other sources of time-varying, subnational data to gauge the impact of Chinese development finance on economic, social, environmental, and governance outcomes.
4. Augment the MBDC methodology to more systematically capture
“unofficial” flows from China to Africa.
The methodology that we have employed to track Chinese development finance did not systematically target “unofficial” financial flows from China to Africa, including joint venture projects (with and without Chinese government involvement), foreign direct investment (with and without Chinese government involvement), aid from private and state-owned Chinese corporations, and aid from Chinese non-governmental organizations. Our objective was to track Chinese development finance in African countries. However, we inadvertently identified a large number of these unofficial activities and chose not to discard the data. We instead separated these (incomplete) data from the official development finance records.
Given the enormous yield of unofficial activities that were captured, we hope to augment our methodology to enable more systematic tracking of these activities, as they help provide a more comprehensive and accurate picture of the wide range of financing modalities Beijing uses to support economic development in Africa.
56 For example, Stage One searches were systematically conducted in English, yielding primary sources in English as well as translations of foreign language media sources. In Stage Two, Google searches in English were supplemented with Baidu searches in Mandarin. However, we recognize these searches may have elided other foreign language media outlets providing valuable project information. For instance, English and Mandarin searches revealed only seven projects in Benin worth US$ 49 million in total. Preliminary Factiva searches in French revealed eight additional projects in Benin, for a combined total of at least US$ 40 million. This suggests our initial results are biased against Francophone countries.
The Factiva search French string used was as follows: (Chine or Chinois or Chin*) near5 (Benin or Beninois or Benin* or Porto-Novo or Cotonou) AND (assistance or subvention or prêt or emprunt or concession* or donat* or donneur or donateur or sans intérêt or intérêt or préférentiel or fonds commun or fond or invest* or finance or aide).
5. Collect development finance data for a DAC donor (or donors) using this media-based method.
While AidData’s MBDC methodology was designed to address the challenge of missing data from non-DAC donors, application of media-based data collection methods to a DAC donor (for whom we have official project-level data) would help reveal the biases and shortcomings of our methodology. It is easier to correct for biases or weaknesses when they are known.
6. Adapt the MBDC methodology for other forms of non-DAC development finance data collection.
AidData has employed MBDC methods to collect some preliminary data for development finance activities funded by Saudi Arabia and Venezuela.57 These pilot exercises have yielded promising results. However, refining these methods to ensure that they are broadly
applicable to non-DAC suppliers of development finance will require more time and careful attention to detail and nuance. While AidData researchers created a methodology designed to track aid from multiple donors, our application to the case of China caused us to create particular categories in the official and unofficial sectors that reflected Chinese aid practices.
When AidData applies this method to other non-traditional donors, we are likely to discover additional nuances and variation in flows that are not captured by the method presented
When AidData applies this method to other non-traditional donors, we are likely to discover additional nuances and variation in flows that are not captured by the method presented